Projects

This page is a comprehensive collection of projects I’ve worked on across academia and industry — ranging from major research and industry initiatives to smaller exploratory and course-based projects. Together, they reflect my evolving interests in engineering, AI, and applied problem-solving.

Professional Projects

Eaton (Jul 2022 – Jun 2025)

AI-Based Manufacturing Defect Reduction for Voltage-Dependent Resistors
• Led a data science–driven quality improvement initiative to reduce manufacturing defects in voltage-dependent resistors, working closely with U.S.-based stakeholders
• Developed machine learning models (support vector machine, ensemble model etc.) to predict batch-level defects with ~85% accuracy, enabling proactive quality control and process optimization
• Built GUI-based analytics applications and a centralized SQL database to digitize production-line data and provide real-time visibility into quality KPIs across manufacturing stations
AI-Based Design Optimization for Overcurrent Protection Devices
• Applied AI-based optimization to support U.S.-based engineering teams in refining overcurrent protection device designs, leveraging historical design data to guide decision-making
• Trained a deep learning model achieving ~80% prediction accuracy, reducing design lead time by three months and delivering $1M in cost savings through faster and more informed design iterations
EV Adoption Analysis for Charging Infrastructure Planning
• Built interactive dashboards using Python based on household power consumption data to visualize electric vehicle adoption trends across U.S. counties
• Derived forward-looking EV growth insights from power usage patterns of households, supporting data-driven planning of EV charging infrastructure
Energy Consumption Analytics and Forecasting for Commercial Buildings
• Developed Python-based energy monitoring KPIs (key performance indicators) to analyze real-time power consumption patterns in commercial buildings
• Applied time-series forecasting using XGBoost to predict power demand, enabling timely adjustments in power generation and load planning
Heat Sink Optimization Using Machine Learning
• Applied machine learning–based optimization to improve heat sink designs for power electronics (e.g., inverters), leveraging simulation-generated thermal performance data
• Trained an XGBoost regression model achieving <5% MAPE (mean absolute percentage error), and used model-driven optimization to identify optimal design configurations for improved thermal efficiency

M.Tech Projects

Machinery Fault Diagnosis Using Machine Learning | M.Tech Project, Prof. Arun Kumar Samantaray, IIT Kharagpur (May 2021 – Apr 2022)
• Devised a neural network using experimental data of accelerometer to detect bearing faults along with their locations (inner race, outer race, or balls) with 95% accuracy
• Applied signal preprocessing with autoregressive (AR) modeling and spectral kurtosis (SK) to improve fault detection
• Benchmarked multiple diagnostic techniques, establishing envelope analysis with AR and SK as the best for noisy signals
• Formulated a convolutional neural network (CNN) to predict the remaining useful life of turbofan engine for proactive, condition-based maintenance
Program to Predict Failure in Composites | Course Project, Dr. Atul Jain, IIT Kharagpur (Feb 2021 – Apr 2021)
• Developed a Python program to compute the properties of a composite material formed after mixing certain amounts of fibre and matrix
• Predicted the possibility of failure in composite using the developed program, including failure location under externally applied loads
Rotation Calculation in a Skewed Plate Using Python | Course Project, Dr. Jeevanjyoti Chakraborty, IIT Kharagpur (Mar 2021 – Apr 2021)
• Developed a Python program to calculate the rotation of the tip chord of a skewed plate fixed at one side with respect to the root chord
• This skewed plate may be used to represent the swept wing of a high-speed aircraft and therefore can have many practical applications
Estimation of Musculoskeletal Forces Using Inverse Dynamics | Course Project,  Prof. Sanjay Gupta, IIT Kharagpur (Jan 2021 – Mar 2021)
• Analyzed hip joint contact forces during human gait and daily activities using inverse dynamics and the HIP98 biomechanical database, focusing on walking and stair-climbing motions
• Developed a lower-limb link-segment model incorporating anthropometric data, ground reaction forces, and kinematic inputs to estimate joint reaction forces and moments non-invasively.
• Implemented Python-based data processing and curve fitting to compute velocity and acceleration of limb segments, validating analytical results against experimentally measured implant data

Self-Projects

Deep Learning for Dog Breed Classification
• Built an end-to-end dog breed classification system using TensorFlow and transfer learning, capable of identifying 120 distinct dog breeds from images
• Performed dataset preprocessing and curation, including image extraction, cleaning, resizing, and formatting to prepare large-scale data for model training
• Experimented with multiple deep learning architectures and configurations, achieving 87% classification accuracy using the EfficientNetV2 convolutional neural network
SkimLit: Rapid Skimming of Medical Literature Using Deep Learning
• Utilised TensorFlow and Keras to develop 1-dimensional convolutional neural network (conv1D) to classify sentences within medical research abstracts into their respective roles (e.g., objective, methods, results)
• Preprocessed and modeled the PubMed 20k RCT dataset, performing text cleaning, vectorization, and embedding to enable efficient training on large-scale biomedical text
Food Vision Big™: Deep Learning Image Classification
• Built and trained a deep learning image classification model using TensorFlow and transfer learning on the Food-101 dataset, scaling from small subsets to full 101 food classes with improved accuracy
• Implemented data preprocessing, batching, mixed-precision training, and model fine-tuning with EfficientNet architectures, and evaluated model performance using TensorBoard and prediction visualization
Spam Classification Using Support Vector Machines
• Built an email spam classifier using a linear Support Vector Machine (SVM), training on ~4,000 labeled emails and evaluating performance on a held-out test set
• Implemented text preprocessing and feature extraction (tokenization, stemming, normalization, vectorization), achieving 99.85% training accuracy and 98.80% test accuracy
Hand-Written Digit Recognition Using Neural Networks
• Developed a three-layer neural network to recognize hand-written digits from 20×20 grayscale images, training the model on 5,000 labeled examples
• Implemented forward propagation and backpropagation for model training, achieving ~99% accuracy on the training dataset and ~98% accuracy on the test dataset

B.Tech Projects

Flow Boiling in Microchannels | B.Tech Project, Prof. Manabendra Pathak, IIT Patna (Jul 2019 – May 2020)
• Designed and fabricated a diverging-depth microchannel heat sink to achieve high heat transfer rates in compact, confined spaces, with a focus on electronics cooling applications
• Performed flow boiling experiments under high heat flux conditions, demonstrating improved heat transfer performance and reduced flow instabilities compared to conventional straight microchannels
Design of Capillary Tube for Vapour Compression Systems | Course Project, Dr. Mohd. Kaleem Khan (Sep 2019 – Nov 2019)
• Developed a MATLAB-based computational tool to determine capillary tube design parameters for vapour compression refrigeration systems based on operating requirements and selected refrigerant properties
• Implemented thermodynamic and fluid-flow calculations to estimate tube length and diameter, enabling systematic sizing of capillary tubes for different refrigeration scenarios
Whiteboard Cleaner | Course Project, Dr. Rishi Raj, Dr. Atul Thakur, Dr. Anirban Bhattacharya, IIT Patna (Aug 2018 – Apr 2019)
• Designed and fabricated a lightweight, low-cost mechanical whiteboard cleaning device to reduce manual effort and time required in wiping the whiteboards
• Incorporated an adjustable screw-based pressure mechanism, allowing effective cleaning across different board conditions and duster wear levels
Atmospheric Water Harvesting Using Biomass Gasification | Dr. Rishi Raj, IIT Patna (May 2018 – Jul 2018)
• Worked on the development of a biomass gasification–based atmospheric water harvester (AWH) aimed at extracting potable water from humid air for water-scarce regions
• Optimized system efficiency by analyzing pre-cooling conditions and heat exchanger design, using MATLAB-based thermodynamic modeling to predict air and water outlet conditions under varying operating inputs
Structural Health Monitoring Using Electromechanical Impedance (EMI) Technique | Dr. Mayank Tiwari, IIT Patna (Feb 2018 – Apr 2018)
• Developed a prototype for non-destructive structural health monitoring of civil infrastructure such as bridges and buildings using the electromechanical impedance (EMI) method
• Utilized PZT (lead zirconate titanate) piezoelectric transducers to detect changes in mechanical impedance, enabling early identification of structural damage without invasive testing
Low-Cost Water Purification for Rural Communities | Dr. Papia Raj, IIT Patna (Feb 2018 – Apr 2018)
• Conducted field surveys in Bihta village to understand local water quality issues, user constraints, and prevalent impurities through direct interaction with community members
• Translated field data into a practical design manual for developing a low-cost, context-specific water purification solution, tailored to local water conditions and affordability constraints